Virt-Manager – KVM Not Available

arch linuxkvmlibvirtqemu

I'm running Arch Linux and have been running into trouble trying to use QEMU/KVM on virt-manager, and this problem is likely to address my issues while using the similar gnome-boxes.

Resolution

This error was related to the driver opencl-amdgpu-pro-orca. This answer details some background of what causes the error. Here's how I went about resolving the problem:

  1. Run sudo systemctl edit libvirtd, which opens Nano.
  • That command creates a temporary file which will save to become override.conf within the directory of the libvirtd service.
  1. Enter the following into this (empty in my case) file:
[Service]
Environment="LD_LIBRARY_PATH=/usr/lib"
  1. Restart libvirtd with sudo systemctl restart libvirtd.
  2. Open virt-manager. The error should not be present.
  • It's worth noting that I did not need to reboot following these changes.

The rest of the question as it originally was is detailed below.

Problem

Upon opening virt-manager, a connection appears to be successfully made with QEMU/KVM, and I am given the option to create a new VM. When I attempt to do so, I experience this error:

Warning: KVM is not available. This may mean the KVM package is not installed, or the KVM kernel modules are not loaded. Your virtual machines may perform poorly.

Log

journalctl -f returns this when I reproduce the error:

Oct 14 22:23:00 okcomputer libvirtd[35615]: internal error: Failed to start QEMU binary /usr/bin/qemu-system-x86_64 for probing: /usr/bin/qemu-system-x86_64: symbol lookup error: /usr/lib/libvirglrenderer.so.1: undefined symbol: gbm_bo_unmap
Oct 14 22:23:00 okcomputer libvirtd[35615]: Failed to probe capabilities for /usr/bin/qemu-system-x86_64: internal error: Failed to start QEMU binary /usr/bin/qemu-system-x86_64 for probing: /usr/bin/qemu-system-x86_64: symbol lookup error: /usr/lib/libvirglrenderer.so.1: undefined symbol: gbm_bo_unmap

Research and attempts at resolution

Arch Wiki

This is an issue that seems quite common and I have researched extensively. I initially followed the Arch Wiki, reading the KVM, QEMU, and libvirt pages, in that order.

KVM

  1. My processor does have support for virtualisation and it is enabled; LC_ALL=C lscpu | grep Virtualization outputs:

LC_ALL=C lscpu | grep Virtualization

Virtualisation support is enabled in my BIOS – both VT-x and VT-d.

  1. This is where the issue may lie – kernel modules. Running zgrep CONFIG_KVM /proc/config.gz outputs the following:
CONFIG_KVM_GUEST=y
# CONFIG_KVM_DEBUG_FS is not set
CONFIG_KVM_MMIO=y
CONFIG_KVM_ASYNC_PF=y
CONFIG_KVM_VFIO=y
CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT=y
CONFIG_KVM_COMPAT=y
CONFIG_KVM=m
CONFIG_KVM_INTEL=m
CONFIG_KVM_AMD=m
CONFIG_KVM_AMD_SEV=y
CONFIG_KVM_MMU_AUDIT=y

lsmod | grep kvm outputs this:

kvm_intel             311296  0
kvm                   778240  1 kvm_intel
irqbypass              16384  1 kvm

I noticed that kvm_intel seems to be set to zero, but is mentioned at the end of the following line. I'm not particularly familiar with how kernel modules work, to be clear, but I decided to try load the module with sudo modprobe kvm_intel, which returns nothing – a good sign, as far as I'm aware. Running sudo modprobe kvm to be certain is the same.

QEMU

  1. I installed QEMU like any other package.

libvirt

  1. I installed libvirt and virt-manager like any other package.
  2. I ensured I had both ebtables and dnsmasq installed and up-to-date.
  3. I created /etc/polkit-1/rules.d/50-libvirt.rules and added this to the file:
/* Allow users in kvm group to manage the libvirt
daemon without authentication */
polkit.addRule(function(action, subject) {
    if (action.id == "org.libvirt.unix.manage" &&
        subject.isInGroup("kvm")) {
            return polkit.Result.YES;
    }
});
  1. I added myself to the kvm group (and libvirt, which I think I saw somewhere online?) via useradd; this can be confirmed by running groups <username>, which I did:

wheel kvm libvirt <username>

  1. I then enabled and started the libvirtd and virtlogd services, and rebooted.

  2. I then tested whether virsh works in addition to virt-manager; virsh outputs the following:

Welcome to virsh, the virtualization interactive terminal.

Type:  'help' for help with commands
       'quit' to quit

virsh #

Additional research

As I mentioned this is a seemingly widespread issue, but following the instructions of many a Stack Exchange answer results in generally the same thing.

I have tested the following:

Countless other articles and answers offer similar advice.

Software specifications

  • Kernel 5.3.5-arch1-1-ARCH (latest default Arch kernel)
  • libvirt 5.6.0-1
  • virt-manager 2.2.1-1
  • QEMU 4.1.0-2
  • polkit 0.116-3

I have no idea if video-related packages are relevant, but the proprietary drivers have proven to be problematic in the past.

  • mesa 192.2.1-1
  • xf86-video-amdgpu 19.0.1-1
  • opencl-amdgpu-pro-orca 19.30_855429-1

Hardware specifications

  • CPU Intel i7 4790k
  • Motherboard Asus Maximus VII Hero

Will provide any necessary additional specs as requested.

Best Answer

Your problem is: opencl-amdgpu-pro-orca.
It delivers its own libgbm.so.1 which is missing that symbol even with latest AUR version with a default. It is put in search path first.
So You can disable these drivers or try to get / compile a fixed / updated version that includes this feature - or try to disable that lib search path only for the libvirtd servcie (add LD_LIBRARY_PATH=/usr/lib environment variable to service startup procedure).

How to analyze such issues for example:

ldd /usr/lib/libvirglrenderer.so.1
ldd /usr/lib/libvirglrenderer.so.1|awk '$3!=""{print $3}'|while read file; do objdump -TC -j.text $file|grep gbm_bo_unmap && echo $file; done
LD_LIBRARY_PATH=/usr/lib ldd /usr/lib/libvirglrenderer.so.1|awk '$3!=""{print $3}'|while read file; do objdump -TC -j.text $file|grep gbm_bo_unmap && echo $file; done
Related Question